Collision Free Path Planning for Underwater Vehicles in Rapidly Changing Environments

This paper presents an obstacle avoidance path planning algorithm designed to generate smooth paths for underwater robotic systems that operate in dynamic environments. Using the kinematics of the system, an initial path is generated which is further optimized considering the constraints of the system and the environment. The correlation between path states is embedded into a kernel used throughout the optimization. This produces a more informative optimization process that leads to changes in one state based on all other states. However, the use of this correlation between path states may lead to an exhaustive computational effort for highly dimensional systems. Therefore, the proposed approach, named AmaxGPMP, introduces a strategy capable of reducing the needed information to develop these kernels while accurately describing the correlation among states, hence decreasing the computation time. The proposed path planner was tested in simulation and experimentally on a BlueROV2 Heavy vehicle that was modified to enable autonomous capabilities. The results demonstrate the ability of AmaxGPMP to successfully generate smooth, feasible, and safe behaviors for autonomous underwater vehicles.